Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging...

83
1 Ministry of Natural Resources Murray Woods Murray Woods Ontario Ministry of Natural Resources Ontario Ministry of Natural Resources Southern Science and Information Southern Science and Information North Bay North Bay [email protected] [email protected] Semi Semi - - Automated Imagery Analysis Automated Imagery Analysis and LiDAR Enhancements to and LiDAR Enhancements to Natural Resource Inventories Natural Resource Inventories

Transcript of Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging...

Page 1: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

1

Ministry of Natural Resources

Murray WoodsMurray WoodsOntario Ministry of Natural ResourcesOntario Ministry of Natural Resources

Southern Science and InformationSouthern Science and InformationNorth BayNorth Bay

[email protected]@Ontario.ca

SemiSemi--Automated Imagery AnalysisAutomated Imagery Analysisand LiDAR Enhancements toand LiDAR Enhancements toNatural Resource InventoriesNatural Resource Inventories

Page 2: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

2

AAdvanceddvanced FForestorest RResourceesource IInventorynventory TTechnologiesechnologies

Doug Pitt – Don Leckie – François Gougeon – Paul Treitz - Al Stinson – Murray Woods - Dave Nesbitt

Page 3: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

3

Project Objective• push “beyond” traditional forestinventory descriptors• leverage existing projects currentlyunderway• advance existing methods andapproaches• link multiple remote-sensing techniquesto derive products at multiple scales

(Tree → Stand → Landscape)

• move toward linking solid inventoryattributes and wood fibre characteristics• develop tools to transfer science &research efforts to government and theprivate sector• ‘operationalize’ the methods developed

Page 4: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

4

Project Focus

• Use of LiDAR technologies to enhance thescope and resolution of natural resourceinventories

• Improvement of semi-automated image analysistechniques to classify tree crowns and developtree/stand polygon inventories

• Use independently or fuse these technologies toimprove natural resource inventories.

Page 5: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

5

Explore the potential of:

1. Using LiDAR formeasurement andprediction of stand levelinventory metrics

2. Determine of appropriateacquisition intensities for arange of forest types

Page 6: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

6

What is LiDAR?What is LiDAR?LLiightght DDetectionetection AAndnd RRanginganging

• Active remote sensing technology

• Involves transmitting and receiving~150,000 pulses of laser light persecond

• Pulses strike the surface of theearth and with each pulse get ameasurement of the time and angleof each return

• If a laser pulse hits an objectthrough which it can penetrate, it willproduce range and intensitymeasurements for each surface it hitsresulting in multiple returns

Page 7: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

7

LiDAR Intro 1LiDAR Intro 1x,y,z

x,y,z+range

range

)/(__*2

)(_)( smlightofspeed

stimeelapsedmrange

Page 8: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

8

LiDAR Intro 2LiDAR Intro 2

x+x1,y,z+z1

z1

x1

θ

range

x,y,z

x,y,z+range

range

Page 9: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

9

Example of flight line RAW LiDAR stripExample of flight line RAW LiDAR strip

Page 10: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

10

Ground Hits OnlyGround Hits Only

NonNon--Ground Hits OnlyGround Hits Only

Classification of ReturnsClassification of Returns

Page 11: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

11

RGB

Page 12: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

12

DSM

RGB

Page 13: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

13

DSM

DTM

RGB

Page 14: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

14

DSM

DTM

RGB

Profile

Page 15: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

15

Digital Surface Model (DSM)Digital Surface Model (DSM)

Digital Terrain Model (DTM)Digital Terrain Model (DTM)

Canopy Height Model (CHM)Canopy Height Model (CHM)

From St-Onge, B., Treitz, P., Wulder, M., Kurtz, W., Gillis,M. 2004. Retrospective mapping of structural andbiomass changes in forest ecosystems usingphotogrammetryand laser altimetry,AmercianGeophysical Union/Canadian Geophysical Union JointAssembly, Montreal, May 17-21.

Page 16: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

16

Quick and Easyproduction of raster-based data productsfrom lidar data..

• Digital surface models• Digital terrain models• Canopy height model• Point density grids• Canopy closure grids• Height structure index grids

Page 17: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

17

• Detailed Surface Models– Digital Surface Models (DSM)– Digital Terrain Models (DTM)– Canopy Height Models (CHM)

LiDAR’s Contribution to Precision Forest Inventory

DTMDTM

DSMDSM

CHMCHM

Page 18: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

18

LiDAR’s Contribution to Precision Forest Inventory• Detailed Surface Models

– Digital Surface Models– Digital Terrain Models– Canopy Height Models

• Detailed Digital Terrain Model– Supporting

• Identifying surficial geologyValue A

ddedV

alue Added

Eskers Dunes

Used with permission from Al Thorne, Tembec.Used with permission from Al Thorne, Tembec.

Page 19: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

19

Northern Ontario 20m DTMNorthern Ontario 20m DTM

LiDAR provides Improved Digital Terrain Models

Page 20: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

20

LiDAR derived 5m DTMLiDAR derived 5m DTM

LiDAR provides Improved Digital Terrain Models

Page 21: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

21

OBM 10m DEM 2.0

LiDAR 1m DEM

Page 22: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

22

LiDAR’s Contribution to Precision Forest Inventory• Detailed Surface Models

– Digital Surface Models– Digital Terrain Models– Canopy Height Models

• Detailed Digital Terrain Model– Supporting

• Identifying surficial Geology• Hydrological modellingV

alue Added

Value A

dded

Predictive Hydrology Models

Predictive Streams

Predicted Drainage requiring Culvert installationPredicted Drainage requiring Culvert installation

•• 86% reliable86% reliable predictions either right on or withinpredictions either right on or within 15 meters15 meters ofofpredicted streams.predicted streams.

•• 95% reliable95% reliable predictions either right on or withinpredictions either right on or within 35 meters35 meters ofofpredicted streams.predicted streams.

OBM Mapped Stream Locations areOBM Mapped Stream Locations are……““completely unreliable for block layout and water crossingscompletely unreliable for block layout and water crossings””

Mark Joran, Millson Forestry ServiceMark Joran, Millson Forestry Service

Page 23: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

23

LiDAR’s Contribution to Precision Forest Inventory• Detailed Surface Models

– Digital Surface Models– Digital Terrain Models– Canopy Height Models

• Detailed Digital Terrain Model– Supporting

• Identifying surficial Geology• Hydrological modelling• Wetland classification

Value A

ddedV

alue Added

Improved Predictive Wetland Classification in N.Improved Predictive Wetland Classification in N.Ontario when compared to current 20m DEMOntario when compared to current 20m DEM

•• ““LiDAR had problemsLiDAR had problemswith obtaining barewith obtaining bareearth returns in alderearth returns in alderthickets and cedarthickets and cedarwetlandswetlands””

Adam HoggAdam Hogg -- IMAIMA

Page 24: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

24

LiDAR’s Contribution to Precision Forest Inventory• Detailed Surface Models

– Digital Surface Models– Digital Terrain Models– Canopy Height Models

• Detailed Digital Terrain Model– Supporting

• Identifying surficial Geology• Hydrological modelling• Wetland identification• Predictive ELC

Value

Value

Added

Added

Output from LandMapR

Page 25: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

25

• Detailed Surface Models– Digital Surface Models– Digital Terrain Models– Canopy Height Models

• Detailed Digital Terrain Model– Supporting

• Identifying surficial Geology• Hydrological modelling• Wetland identification• Predictive ELC

Value

Value

Added

Added

LiDAR’s Contribution to Natural Resource Inventories

Predicted Ecosite MappingProviding additional information:• vegetation communities• Soil depth,• Soil texture• Soil moisture regime,• Soil nutrients• Landform•Geomorphology descriptors (sand,gravel, rock, fluvial, alluvial fans,lacustrine, glaciofluvial, etc…).

Page 26: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

26

LiDAR’s Contribution to Precision Forest Inventory• Detailed Surface Models

– Digital Surface Models– Digital Terrain Models– Canopy Height Models

• Detailed Digital Terrain Model– Supporting

• Identifying surficial Geology• Hydrological modelling• Wetland identification• Predictive Ecosystem mapping• Operational considerations

– road construction– skid trail layout– water crossings

Value A

ddedV

alue Added

Virtual Road layoutVirtual Road layout

Origin

Destination

DEM 2m

Profile from DEM

Wetness index and trail layoutWetness index and trail layout

FPInovations Ferric

Page 27: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

27

LiDAR’s Contribution to Precision Forest Inventory• Detailed Surface Models

– Digital Surface Models– Digital Terrain Models– Canopy Height Models

• Detailed Digital Terrain Model– Supporting

• Identifying surficial Geology• Hydrological modelling• Wetland identification• Predictive Ecosystem mapping• Operational considerations

– road construction– skid trail layout– water crossings

• Direct measurement of:• Stand/Tree Heights• Crown Closure

Value A

dded

1m Canopy Height Model1m Canopy Height Model

Canopy Closure Comparison

0

10

20

30

40

50

60

70

80

90

100

1 2 3 4 5 6

Plot

0

10

20

30

40

50

60

70

80

90

100

Calculated Stand Top H

eight (m)

Lidar Maximum Vegetation Return vs. Calculated Top Height

0

5

10

15

20

25

30

35

40

0 5 10 15 20 25 30 35 40

Lidar Max Vegetation Return (m)

Mean - -0.79m

Min - -4.85m

Max - 4.07m

Std - 1.50m

Page 28: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

28

LiDAR’s Contribution to Precision Forest Inventory• Detailed Digital Surface Models

– Digital Surface Models– Digital Terrain Models– Canopy Height Models

• Detailed Digital Terrain Model– Supporting

• Identifying surficial Geology• Hydrological modelling• Wetland identification• Predictive Ecosystem mapping• Operational considerations

– road construction– skid trail layout– water crossings

• Direct measurement of:• Stand/Tree Heights• Crown Closure

• Individual Tree Information– Tree crown centre– Tree crown radii– Tree height

Value A

ddedV

alue Added

ITC Suite – François GougeonTreeVaw - Sorin Popescu

Page 29: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

29

LiDAR’s Contribution to Precision Forest Inventory• Detailed Digital Surface Models

– Digital Surface Models– Digital Terrain Models– Canopy Height Models

• Detailed Digital Terrain Model– Supporting

• Identifying surficial Geology• Hydrological modelling• Wetland identification• Predictive Ecosystem mapping• Operational considerations

– road construction– skid trail layout– water crossings

• Direct measurement of:• Stand/Tree Heights• Crown Closure

• Individual Tree Information

• Statistically based predictions ofstand attributes

Value A

ddedV

alue Added

Page 30: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

30

Lidar-basedmodels of

forestparameters

• Height• Volume (GTV, GMV)• Basal area• Density• Quadratic mean DBH• Biomass• Diameter and Basal

Area Distributions

Page 31: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

31

Canopy Height Model (CHM)

Page 32: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

32

LiDAR Predictors

• Statistical– Mean, Standard Deviation,

Absolute Deviation, Skew, Kurtosis

• Percentiles– Deciles (p10 … p90) and Maximum Height

• Density– d1 … d9

– Da : Number of first returns divided by all returns.

– Db : Number of first returns classified as non-ground divided by all returns.

• derived from First, or All returns; with or without a z threshold

Page 33: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

33

P0

heig

ht

1

q(ht)

q(ht)

Concept of Canopy Height Metrics

0

5

10

15

20

25

30

4.2966e+54.2967e+5

4.2968e+54.2969e+5

4.2970e+54.2971e+5

4.681400e+64.681405e+64.681410e+64.681415e+64.681420e+64.681425e+64.681430e+64.681435e+6

Z Data

X Data

Y Data

ACFL

9 values that divide sorted data into 10 equal parts with each part representing1/10th of the sample or population

Page 34: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

34

Concept of Canopy Density Metrics

0

5

10

15

20

25

30

4.2966e+54.2967e+5

4.2968e+54.2969e+5

4.2970e+54.2971e+5

4.681400e+64.681405e+64.681410e+64.681415e+64.681420e+64.681425e+64.681430e+64.681435e+6

Z Data

X DataY Data

ACFL

P0

heig

ht

1

D(%)

• Range of heights divided into 10 equal intervals.

• Cumulative proportion of returns starting from the lowest interval.

Page 35: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

35

Statistical Analyses

• Linear / Non-Linear Regression

• Model Diagnosis– Test for Normality: Shapiro-Wilks Test– Test for Homoschedasticity: Modified Levene’s Test– Multicollinearity: Variance Inflation Factors (VIF) < 10– Natural Logarithm Transformation

• Validation– Independent data sets or PRESS Procedure

nn xxy ...110

Page 36: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

36

Workflow – Model Building

Page 37: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

37

Decimation 0 ~3 p/mDecimation 0 ~3 p/m22 Decimation 1 ~1.6 p/mDecimation 1 ~1.6 p/m22 Decimation 2 ~0.5 p/mDecimation 2 ~0.5 p/m22

What is the appropriate density to acquireWhat is the appropriate density to acquireLiDAR at for stand level prediction?LiDAR at for stand level prediction?

Page 38: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

38

Decimation 0 ~3 p/mDecimation 0 ~3 p/m22 Decimation 1 ~1.6 p/mDecimation 1 ~1.6 p/m22 Decimation 2 ~0.5 p/mDecimation 2 ~0.5 p/m22

Cumulative Proportion of LiDAR Returnsfor Level of LiDAR Decimation

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

2 4 6 8 10 12 14 16 18 20 22 24 26

Height Class

%D0D1D2

Page 39: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

39

Study Sites

• Three (3) main Ontario study sites:– Swan Lake (SL) Reserve (n=32)– Petawawa Research Forest (PRF) (n=32)– Romeo Mallette (RM) Forest (n=130)

Jack PineJack Pine Black SpruceBlack SpruceGrtGrt LksLks PinePine

TolerantTolerant HwdsHwds IntolerantIntolerant HwdsHwds MixedwoodsMixedwoods

Page 40: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

40

LiDAR DataPeriod: Summer 2007Sensor: Optech ALTM 3100Altitude: 1000 mOverlap: N/ASpeed: 120 knotsSystem PRF: 100 kHzScan Freq: 54 HzScan Half Angle: 13°Cross Track Resolution: 0.499 mDown Track Resolution: 0.572 m

~ 3 point/m2

Page 41: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

41

WorkflowDecimate LiDAR Data~3 pulses/m2

~1.6 pulses/m2

~.5 pulses/m2

Page 42: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

42

Petawawa Research Forest – Grt. Lakes Pine(Pw/Pr/Pj Natural + Planted + Thinned)

Variable Decimation Level 0 Decimation Level 1 Decimation Level 2

R2 RMSE(%)

R2 RMSE(%)

R2 RMSE(%)

SUMBA .89 4.8(13.3)

.89 4.7(13.2)

.88 4.9(13.7)

SUMGTV .93 56.1(13.4)

.94 51.2(12.3)

.92 60.0(14.4)

DENSITY .74 207.8(34.7)

.72 214.9(35.8)

.68 226.9(37.8)

QMDBH .87 3.9(12.7)

.86 4.0(12.9)

.87 3.8(12.3)

AVGHT .94 1.3(5.7)

.94 1.3(6.0)

.94 1.3(5.7)

TOPHT .95 1.2(4.4)

.95 1.2(4.5)

.94 1.3(4.7)

SUMBIO .74 29,818(21.0)

.76 28,743(20.3)

,78 27,572(19.4)

Current StatusPreliminary Results

Page 43: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

43

Variable Decimation Level 0 Decimation Level 1 Decimation Level 2

R2 RMSE(%)

R2 RMSE(%)

R2 RMSE(%)

SUMBA .49 2.7(10.6)

.49 2.7(10.7)

.58 2.4(9.7)

SUMGTV .59 25.4(11.2)

.60 24.9(11.0)

.61 24.8(11.0)

DENSITY .84 42.8(10.5)

.86 39.8(9.8)

.83 43.7(10.7)

QMDBH .69 2.0(7.3)

.72 1.9(7.0)

.70 2.0(7.1)

AVGHT .84 0.6(3.4)

.84 0.6(3.4)

.85 0.6(3.2)

TOPHT .82 0.7(3.0)

.85 0.7(2.8)

.86 0.7(2.7)

SUMBIO .46 24,795(12.6)

.50 23,811(12.1)

.58 23,811(12.1)

Tolerant Hardwoods

Current StatusPreliminary Results~ 3 pts/m2 ~ 1.6 pts/m2 ~ 0.4 pts/m2

Page 44: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

44

Romeo Malette Forest – Mixedwoods

Preliminary ResultsDensity Basal Area

Dbhq GTV Biomass

Mixedwoods - Decimation Level 0

0

10

20

30

40

50

60

70

0 10 20 30 40 50 60 70

Measued BA m2/ha

Pred

icte

d BA

m2/

ha

Mixedwoods - Devimation Level 0

-100

0

100

200

300

400

500

600

700

800

0 100 200 300 400 500 600 700 800

Measured GTV m3/ha

Pred

icte

d G

TV m

3/ha

Mixedwoods - Decimation Level 0

0

5

10

15

20

25

30

35

0 5 10 15 20 25 30 35

Measured Dbhq cm

Pred

icte

d D

bhq

cm

Mixedwoods - Decimation Level 0

0

50000

100000

150000

200000

250000

300000

350000

0 50000 100000 150000 200000 250000 300000 350000

Measured Biomass Kg/ha

Pred

icte

d Bi

omas

s Kg

/ha

Top Height

Mixedwoods - Decimation Level 0

0

5

10

15

20

25

30

35

40

0 5 10 15 20 25 30 35 40

Measured Top Height m

Pred

icte

d To

p H

eigh

t m

Mixedwoods - Decimation Level 0

0

500

1000

1500

2000

2500

3000

0 500 1000 1500 2000 2500 3000

Measured Density Stems/ha

Pred

icte

d D

ensi

ty S

tem

s/ha

Page 45: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

45

Romeo Malette Forest – MixedwoodsVariable Decimation Level 0 Decimation Level 1 Decimation Level 2

R2 RMSE(%)

R2 RMSE(%)

R2 RMSE(%)

SUMBA .72 5.5(16.8)

.73 5.3(16.4)

.72 5.5(16.8)

SUMGTV .82 48.2(18.2)

.83 47.1(17.8)

.82 48.2(18.2)

DENSITY .12 309.4(28.1)

.11 310.9(28.2)

.12 309.4(28.1)

QMDBH .61 1.6(8.5)

.62 1.6(8.4)

.61 1.6(8.5)

AVGHT .78 0.9(6.0)

.82 0.8(5.3)

.78 0.9(6.0)

TOPHT .97 0.6(2.8)

.97 0.6(2.8)

.97 0.6(2.8)

SUMBIO .75 22,510(16.2)

.81 19,469(14.1)

.75 22,510(16.25)

Current Status

This analysis: All Returns – No Z Threshold imposed

Preliminary Results

Page 46: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

46

Applying Models to the LandscapeLiDAR Predictor Surfaces

•Each surfacecorresponds to a LIDARpredictor

•Example is 12,000 ha ofPetawawa ResearchForest

Page 47: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

47

Image based Interpretation

Page 48: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

48

Image based Interpretation – Polygon Mask

Page 49: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

49

LiDAR derived Canopy Height Model

Page 50: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

50

Operational LiDAR Enhancements

Stratify by:

•Forest type

•Forest unit

•etc…

Page 51: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

51

Operational LiDAR Enhancements

Not to scale – illustrative purposes only

400m400m22 Predictive Grid CellsPredictive Grid Cells

Page 52: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

52

Operational LiDAR EnhancementsAverage Height (m) 400mAverage Height (m) 400m22 RasterRaster

Page 53: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

53

Operational LiDAR EnhancementsAverage Height (m) 400mAverage Height (m) 400m22 RasterRaster

Page 54: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

54

Operational LiDAR EnhancementsAverage Height (m) 400mAverage Height (m) 400m22 RasterRaster

Page 55: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

55

Operational LiDAR EnhancementsBasal Area (mBasal Area (m22/ha) 400m/ha) 400m22 RasterRaster

Page 56: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

56

Operational LiDAR EnhancementsGTV (mGTV (m33/ha) 400m/ha) 400m22 RasterRaster

Dbhq (cm) 400mDbhq (cm) 400m22 RasterRaster

Page 57: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

57

Diameter Distribution Modelling

From: Quantifying diameter distributions of uneven-aged toleranthardwood stands using LiDAR

Jones, T, Woods, M. and Lim K 2009 Inpress Silvilaser 2009

Sawtimber size class

pole small medium large

Basal area (m2 ha -1)

0

2

4

6

8

10

Observed basal areaLiDAR predicted basal area

a

b

aa

aa

aa

SLFRR validation plots

0.00

0.05

0.10

0.15

0.20

0.25

0.00

0.05

0.10

0.15

0.20

0.25

0 20 40 60 800.00

0.05

0.10

0.15

0.20

0.25

0.00

0.05

0.10

0.15

0.20

0.25Observed diameter distributionLiDAR estimated

0.00

0.05

0.10

0.15

0.20

0.25

0 20 40 60 800.00

0.05

0.10

0.15

0.20

0.25

Relative Frequency

Relative Frequency

Diameter class (cm)

Validation Data

Page 58: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

58

All Rondeau Vegegation PlotsComparing LiDAR against Ground Measured Heights (Bird Monitoring Classes)

0

5

10

15

20

25

30

0-50 .51-1.30 1.31-3.0 3.1-6.0 6.1-10.0 10.1-20.0 20+

Height Class

Heig

ht m First Returns

All_ReturnsMeasured HT

LiDAR unlocking Vertical Structure for Habitat ModelingLiDAR unlocking Vertical Structure for Habitat Modeling

Percent Vegetation Returns

0 5 10 15 20 25 30 35 40 45

.51-1.3

1.31-3.0

3.0-6.0

6-10

10.1-20

20+

101C

Percent Vegetation Returns

0 5 10 15 20 25 30 35 40 45

.51-1.3

1.31-3.0

3.0-6.0

6-10

10.1-20

20+

105A

% Vegetation Cover by Height Class

Canopy Height Model Sliced Canopy Height Model

Foliage Height Diversity

19

0

5

10

15

20

25

430015 430020 430025 430030 430035 430040 430045 430050 430055 430060

Z

102C

0

5

10

15

20

25

30

430285 430290 430295 430300 430305 430310 430315 430320 430325 430330

Z

Vertical Distribution Ratio = 0.64Vertical Distribution Ratio = 0.64Coefficient of Variation = 56.19Coefficient of Variation = 56.19

Vertical Distribution Ratio = 0.51Vertical Distribution Ratio = 0.51Coefficient of Variation = 39.95Coefficient of Variation = 39.95

Page 59: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

59

The Role of MultibandThe Role of Multiband Orthophotography in theOrthophotography in theproduction of Semiproduction of Semi--Automated Enhanced ForestAutomated Enhanced Forest

Inventories in the Great Lakes St. Lawrence ForestInventories in the Great Lakes St. Lawrence Forest

Enhanced Forest Productivity FundEnhanced Forest Productivity Fund

Murray Woods, Dave NesbittMurray Woods, Dave NesbittSouthern Science & InformationSouthern Science & Information

FranFranççoisois Gougeon, Don LeckieGougeon, Don LeckieCanadian Forest ServiceCanadian Forest Service

Paul Courville, Sue PickeringPaul Courville, Sue PickeringForest Research PartnershipForest Research Partnership

Page 60: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

60

Project Focus

Explore the potential of:• Using high-resolution multi-band

digital imagery for automatedinventory production– Stand Level– Tree Level

Page 61: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

61

Working with Drs GougeonWorking with Drs Gougeonand Leckie at CFSand Leckie at CFS-- PFCPFC ––application evolutionapplication evolution

Working with SilvatechWorking with SilvatechConsulting Ltd.Consulting Ltd. ––operational applicationoperational application

RESEARCH

OPERATIONS

SemiSemi--automatedautomated -- Individual Tree Crown ClassificationIndividual Tree Crown Classification

Page 62: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

62

SemiSemi--automatedautomated -- Individual Species ClassificationIndividual Species Classification

Page 63: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

63

SemiSemi--automatedautomated -- Individual Species ClassificationIndividual Species Classification

RGB

Spectral-Based Masking

Lidar-Based Masking

Page 64: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

64

SemiSemi--automatedautomated -- Individual Species ClassificationIndividual Species Classification

RGB

Spectral-Based Masking

Page 65: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

65

SemiSemi--automatedautomated -- Individual Species ClassificationIndividual Species Classification

RGB

Spectral-Based Masking

Page 66: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

66

SemiSemi--automatedautomated -- Individual Species ClassificationIndividual Species Classification

Page 67: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

67

SemiSemi--automatedautomated -- Individual Species ClassificationIndividual Species Classification

Page 68: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

68

SemiSemi--automatedautomated -- Individual Species ClassificationIndividual Species Classification

RGB

Spectral-Based Masking

Ground trainingGround training

Page 69: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

69

SemiSemi--automatedautomated -- Individual Species ClassificationIndividual Species Classification

RGB

Spectral-Based Masking

Soft copy trainingSoft copy training

Page 70: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

70

SemiSemi--automatedautomated -- Individual Species ClassificationIndividual Species Classification

RGB

Spectral-Based Masking

Green and Blue by species and 0.7 SD of Green

Pw im

Pw o

Pw br

Mh

Ms

Bwd

Bw

AtAtpq

Alt

By

La

Pr

Sw

Sb

Pj BD

Bf

110

120

130

140

150

160

170

130 140 150 160 170 180

Intensity of Green

Inte

nsity

of B

lue

Pw imPw oldPw brMhMsBwdBwAtAtpqAltByLaPrSwSbPjBDBf

nIR and Red by species and 0.7 SD of nIR

Pw im

Pw old

Pw br

Mh

Ms

Bwd

Bw

At

At p Alt

By

La

Pr

Sw

Sb

Pj

BDBf

120

130

140

150

160

170

160 170 180 190 200 210 220

Intensity of near Infrared

Inte

nsity

of R

ed

Pw imPw oldPw brMhMsBwdBwAtAtpqAltByLaPrSwSbPjBDBf

Species Separability

Page 71: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

71

nIR and Red by species and 0.7 SD of nIR

Pw im

Pw old

Pw br

Mh

Ms

Bwd

Bw

At

At p Alt

By

La

Pr

Sw

Sb

Pj

BDBf

120

130

140

150

160

170

160 170 180 190 200 210 220

Intensity of near Infrared

Inte

nsity

of R

ed

Pw imPw oldPw brMhMsBwdBwAtAtpqAltByLaPrSwSbPjBDBf

ITC Species Signatures in Multispectral Space for PRF

Page 72: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

72

SemiSemi--automatedautomated -- Individual Species ClassificationIndividual Species Classification

Page 73: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

73

SemiSemi--automatedautomated -- Individual Species ClassificationIndividual Species Classification

Page 74: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

74

SemiSemi--automatedautomated -- Individual Species ClassificationIndividual Species Classification

Pw Mh Ms Bw As La Pr Sw Sb PjPw 54.19 26.23 0.00 0.00 0.00 12.44 8.87 0.78 3.83 2.17Mh 2.86 66.12 0.00 0.00 0.00 0.46 0.00 0.52 0.18 0.00Ms 2.66 0.00 66.67 0.00 9.24 0.00 9.61 0.00 0.00 0.00Bw 0.00 0.00 0.00 96.77 9.24 0.00 0.18 0.00 0.00 0.00As 3.27 0.55 33.33 3.23 75.63 5.53 6.47 0.00 0.00 0.00La 14.72 3.83 0.00 0.00 0.00 73.27 0.55 0.00 1.28 0.00Pr 11.25 0.55 0.00 0.00 4.20 4.15 68.21 2.08 0.00 4.35Sw 0.41 0.00 0.00 0.00 0.00 0.00 0.92 67.79 5.28 2.17Sb 6.75 0.00 0.00 0.00 0.00 0.46 0.00 10.91 83.97 1.09Pj 2.45 0.00 0.00 0.00 0.00 0.92 3.51 15.06 2.19 86.96Un 1.43 2.73 0.00 0.00 1.68 2.76 1.66 2.86 3.28 3.26

# ITCs 489 183 18 31 119 217 541 385 549 92

Confusion Matrix Results for PRF

Overall accuracy (precision) of 70.4%

Page 75: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

75

SemiSemi--automatedautomated -- Individual Species ClassificationIndividual Species Classification

Page 76: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

76

SemiSemi--automatedautomated -- Individual Species ClassificationIndividual Species Classification

Page 77: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

77

SemiSemi--automatedautomated -- Individual Species ClassificationIndividual Species Classification

Page 78: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

78

SemiSemi--automatedautomated -- Individual Species ClassificationIndividual Species Classification

Page 79: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

79

SemiSemi--automatedautomated -- Individual Species ClassificationIndividual Species Classification

Automated Polygon DelineationAutomated Polygon Delineationwith Species Classificationwith Species Classification

49 Pw 21 Sw 9 Bf 8 Sb 7 Bd 3 Pr 3 Ce31m

42% Crown Closure238 Stems

Page 80: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

80

Semi-Automated Imagery Classificationand LiDAR for Inventory Production

• Semi-automated imagery classification is a maturing area –focus of ongoing national research efforts with CWFC

• Boreal forest species currently best opportunity to benefitfrom imagery analysis techniques– Another “tool” for an interpreter– Objective vs. Subjective approach– Able to process large areas very quickly– Ability to improve and recompile– Driving towards tree level inventories– Hardwood aggregation or over splitting

receiving attention

Romeo Malette Forest Example

Page 81: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

81

Semi-Automated Imagery Classificationand LiDAR for Inventory Production

• LiDAR methods well understood and ready to operationallyimplement– Become an affordable technology– Demonstrating that higher sampling point densities are not required

for better estimates of forest inventory variables.– Current inventory efforts on a 650K ha NE Ontario Forest– Planned inventory project for 1.3 Million ha forest in 2010– Need to continue to expand the areas in which LiDAR has the

potential to help add value to inventories – eg. Stream prediction,forest habitat modeling, renewable energy modeling

Page 82: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

82

In Conclusion, I believe that…These technologies offergreat potential to improvecurrent methods ofinventory production.Sound sustainablemanagement andbusiness decisionsrequire detailed, highquality & spatially explicitresource inventoryinformation…we can’tkeep putting it off.

Page 83: Semi-Automated Imagery Analysis and LiDAR …€¦ · 6 What is LiDAR? Light Detection And Ranging • Active remote sensing technology • Involves transmitting and receiving ~150,000

83

Ministry of Natural Resources

Author: Mike Baldwin

Thank You

Questions

Contact

Murray Woods at:

[email protected]